1 /* 2 * Network-device interface management. 3 * 4 * Copyright (c) 2004-2005, Keir Fraser 5 * 6 * This program is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License version 2 8 * as published by the Free Software Foundation; or, when distributed 9 * separately from the Linux kernel or incorporated into other 10 * software packages, subject to the following license: 11 * 12 * Permission is hereby granted, free of charge, to any person obtaining a copy 13 * of this source file (the "Software"), to deal in the Software without 14 * restriction, including without limitation the rights to use, copy, modify, 15 * merge, publish, distribute, sublicense, and/or sell copies of the Software, 16 * and to permit persons to whom the Software is furnished to do so, subject to 17 * the following conditions: 18 * 19 * The above copyright notice and this permission notice shall be included in 20 * all copies or substantial portions of the Software. 21 * 22 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 23 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 24 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 25 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 26 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 27 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 28 * IN THE SOFTWARE. 29 */ 30 31 #include "common.h" 32 33 #include <linux/kthread.h> 34 #include <linux/sched/task.h> 35 #include <linux/ethtool.h> 36 #include <linux/rtnetlink.h> 37 #include <linux/if_vlan.h> 38 #include <linux/vmalloc.h> 39 40 #include <xen/events.h> 41 #include <asm/xen/hypercall.h> 42 #include <xen/balloon.h> 43 44 #define XENVIF_QUEUE_LENGTH 32 45 46 /* Number of bytes allowed on the internal guest Rx queue. */ 47 #define XENVIF_RX_QUEUE_BYTES (XEN_NETIF_RX_RING_SIZE/2 * PAGE_SIZE) 48 49 /* This function is used to set SKBFL_ZEROCOPY_ENABLE as well as 50 * increasing the inflight counter. We need to increase the inflight 51 * counter because core driver calls into xenvif_zerocopy_callback 52 * which calls xenvif_skb_zerocopy_complete. 53 */ 54 void xenvif_skb_zerocopy_prepare(struct xenvif_queue *queue, 55 struct sk_buff *skb) 56 { 57 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_ENABLE; 58 atomic_inc(&queue->inflight_packets); 59 } 60 61 void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue) 62 { 63 atomic_dec(&queue->inflight_packets); 64 65 /* Wake the dealloc thread _after_ decrementing inflight_packets so 66 * that if kthread_stop() has already been called, the dealloc thread 67 * does not wait forever with nothing to wake it. 68 */ 69 wake_up(&queue->dealloc_wq); 70 } 71 72 int xenvif_schedulable(struct xenvif *vif) 73 { 74 return netif_running(vif->dev) && 75 test_bit(VIF_STATUS_CONNECTED, &vif->status) && 76 !vif->disabled; 77 } 78 79 static bool xenvif_handle_tx_interrupt(struct xenvif_queue *queue) 80 { 81 bool rc; 82 83 rc = RING_HAS_UNCONSUMED_REQUESTS(&queue->tx); 84 if (rc) 85 napi_schedule(&queue->napi); 86 return rc; 87 } 88 89 static irqreturn_t xenvif_tx_interrupt(int irq, void *dev_id) 90 { 91 struct xenvif_queue *queue = dev_id; 92 int old; 93 94 old = atomic_fetch_or(NETBK_TX_EOI, &queue->eoi_pending); 95 WARN(old & NETBK_TX_EOI, "Interrupt while EOI pending\n"); 96 97 if (!xenvif_handle_tx_interrupt(queue)) { 98 atomic_andnot(NETBK_TX_EOI, &queue->eoi_pending); 99 xen_irq_lateeoi(irq, XEN_EOI_FLAG_SPURIOUS); 100 } 101 102 return IRQ_HANDLED; 103 } 104 105 static int xenvif_poll(struct napi_struct *napi, int budget) 106 { 107 struct xenvif_queue *queue = 108 container_of(napi, struct xenvif_queue, napi); 109 int work_done; 110 111 /* This vif is rogue, we pretend we've there is nothing to do 112 * for this vif to deschedule it from NAPI. But this interface 113 * will be turned off in thread context later. 114 */ 115 if (unlikely(queue->vif->disabled)) { 116 napi_complete(napi); 117 return 0; 118 } 119 120 work_done = xenvif_tx_action(queue, budget); 121 122 if (work_done < budget) { 123 napi_complete_done(napi, work_done); 124 /* If the queue is rate-limited, it shall be 125 * rescheduled in the timer callback. 126 */ 127 if (likely(!queue->rate_limited)) 128 xenvif_napi_schedule_or_enable_events(queue); 129 } 130 131 return work_done; 132 } 133 134 static bool xenvif_handle_rx_interrupt(struct xenvif_queue *queue) 135 { 136 bool rc; 137 138 rc = xenvif_have_rx_work(queue, false); 139 if (rc) 140 xenvif_kick_thread(queue); 141 return rc; 142 } 143 144 static irqreturn_t xenvif_rx_interrupt(int irq, void *dev_id) 145 { 146 struct xenvif_queue *queue = dev_id; 147 int old; 148 149 old = atomic_fetch_or(NETBK_RX_EOI, &queue->eoi_pending); 150 WARN(old & NETBK_RX_EOI, "Interrupt while EOI pending\n"); 151 152 if (!xenvif_handle_rx_interrupt(queue)) { 153 atomic_andnot(NETBK_RX_EOI, &queue->eoi_pending); 154 xen_irq_lateeoi(irq, XEN_EOI_FLAG_SPURIOUS); 155 } 156 157 return IRQ_HANDLED; 158 } 159 160 irqreturn_t xenvif_interrupt(int irq, void *dev_id) 161 { 162 struct xenvif_queue *queue = dev_id; 163 int old; 164 bool has_rx, has_tx; 165 166 old = atomic_fetch_or(NETBK_COMMON_EOI, &queue->eoi_pending); 167 WARN(old, "Interrupt while EOI pending\n"); 168 169 has_tx = xenvif_handle_tx_interrupt(queue); 170 has_rx = xenvif_handle_rx_interrupt(queue); 171 172 if (!has_rx && !has_tx) { 173 atomic_andnot(NETBK_COMMON_EOI, &queue->eoi_pending); 174 xen_irq_lateeoi(irq, XEN_EOI_FLAG_SPURIOUS); 175 } 176 177 return IRQ_HANDLED; 178 } 179 180 int xenvif_queue_stopped(struct xenvif_queue *queue) 181 { 182 struct net_device *dev = queue->vif->dev; 183 unsigned int id = queue->id; 184 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, id)); 185 } 186 187 void xenvif_wake_queue(struct xenvif_queue *queue) 188 { 189 struct net_device *dev = queue->vif->dev; 190 unsigned int id = queue->id; 191 netif_tx_wake_queue(netdev_get_tx_queue(dev, id)); 192 } 193 194 static u16 xenvif_select_queue(struct net_device *dev, struct sk_buff *skb, 195 struct net_device *sb_dev) 196 { 197 struct xenvif *vif = netdev_priv(dev); 198 unsigned int size = vif->hash.size; 199 unsigned int num_queues; 200 201 /* If queues are not set up internally - always return 0 202 * as the packet going to be dropped anyway */ 203 num_queues = READ_ONCE(vif->num_queues); 204 if (num_queues < 1) 205 return 0; 206 207 if (vif->hash.alg == XEN_NETIF_CTRL_HASH_ALGORITHM_NONE) 208 return netdev_pick_tx(dev, skb, NULL) % 209 dev->real_num_tx_queues; 210 211 xenvif_set_skb_hash(vif, skb); 212 213 if (size == 0) 214 return skb_get_hash_raw(skb) % dev->real_num_tx_queues; 215 216 return vif->hash.mapping[vif->hash.mapping_sel] 217 [skb_get_hash_raw(skb) % size]; 218 } 219 220 static netdev_tx_t 221 xenvif_start_xmit(struct sk_buff *skb, struct net_device *dev) 222 { 223 struct xenvif *vif = netdev_priv(dev); 224 struct xenvif_queue *queue = NULL; 225 unsigned int num_queues; 226 u16 index; 227 struct xenvif_rx_cb *cb; 228 229 BUG_ON(skb->dev != dev); 230 231 /* Drop the packet if queues are not set up. 232 * This handler should be called inside an RCU read section 233 * so we don't need to enter it here explicitly. 234 */ 235 num_queues = READ_ONCE(vif->num_queues); 236 if (num_queues < 1) 237 goto drop; 238 239 /* Obtain the queue to be used to transmit this packet */ 240 index = skb_get_queue_mapping(skb); 241 if (index >= num_queues) { 242 pr_warn_ratelimited("Invalid queue %hu for packet on interface %s\n", 243 index, vif->dev->name); 244 index %= num_queues; 245 } 246 queue = &vif->queues[index]; 247 248 /* Drop the packet if queue is not ready */ 249 if (queue->task == NULL || 250 queue->dealloc_task == NULL || 251 !xenvif_schedulable(vif)) 252 goto drop; 253 254 if (vif->multicast_control && skb->pkt_type == PACKET_MULTICAST) { 255 struct ethhdr *eth = (struct ethhdr *)skb->data; 256 257 if (!xenvif_mcast_match(vif, eth->h_dest)) 258 goto drop; 259 } 260 261 cb = XENVIF_RX_CB(skb); 262 cb->expires = jiffies + vif->drain_timeout; 263 264 /* If there is no hash algorithm configured then make sure there 265 * is no hash information in the socket buffer otherwise it 266 * would be incorrectly forwarded to the frontend. 267 */ 268 if (vif->hash.alg == XEN_NETIF_CTRL_HASH_ALGORITHM_NONE) 269 skb_clear_hash(skb); 270 271 xenvif_rx_queue_tail(queue, skb); 272 xenvif_kick_thread(queue); 273 274 return NETDEV_TX_OK; 275 276 drop: 277 vif->dev->stats.tx_dropped++; 278 dev_kfree_skb(skb); 279 return NETDEV_TX_OK; 280 } 281 282 static struct net_device_stats *xenvif_get_stats(struct net_device *dev) 283 { 284 struct xenvif *vif = netdev_priv(dev); 285 struct xenvif_queue *queue = NULL; 286 unsigned int num_queues; 287 u64 rx_bytes = 0; 288 u64 rx_packets = 0; 289 u64 tx_bytes = 0; 290 u64 tx_packets = 0; 291 unsigned int index; 292 293 rcu_read_lock(); 294 num_queues = READ_ONCE(vif->num_queues); 295 296 /* Aggregate tx and rx stats from each queue */ 297 for (index = 0; index < num_queues; ++index) { 298 queue = &vif->queues[index]; 299 rx_bytes += queue->stats.rx_bytes; 300 rx_packets += queue->stats.rx_packets; 301 tx_bytes += queue->stats.tx_bytes; 302 tx_packets += queue->stats.tx_packets; 303 } 304 305 rcu_read_unlock(); 306 307 vif->dev->stats.rx_bytes = rx_bytes; 308 vif->dev->stats.rx_packets = rx_packets; 309 vif->dev->stats.tx_bytes = tx_bytes; 310 vif->dev->stats.tx_packets = tx_packets; 311 312 return &vif->dev->stats; 313 } 314 315 static void xenvif_up(struct xenvif *vif) 316 { 317 struct xenvif_queue *queue = NULL; 318 unsigned int num_queues = vif->num_queues; 319 unsigned int queue_index; 320 321 for (queue_index = 0; queue_index < num_queues; ++queue_index) { 322 queue = &vif->queues[queue_index]; 323 napi_enable(&queue->napi); 324 enable_irq(queue->tx_irq); 325 if (queue->tx_irq != queue->rx_irq) 326 enable_irq(queue->rx_irq); 327 xenvif_napi_schedule_or_enable_events(queue); 328 } 329 } 330 331 static void xenvif_down(struct xenvif *vif) 332 { 333 struct xenvif_queue *queue = NULL; 334 unsigned int num_queues = vif->num_queues; 335 unsigned int queue_index; 336 337 for (queue_index = 0; queue_index < num_queues; ++queue_index) { 338 queue = &vif->queues[queue_index]; 339 disable_irq(queue->tx_irq); 340 if (queue->tx_irq != queue->rx_irq) 341 disable_irq(queue->rx_irq); 342 napi_disable(&queue->napi); 343 del_timer_sync(&queue->credit_timeout); 344 } 345 } 346 347 static int xenvif_open(struct net_device *dev) 348 { 349 struct xenvif *vif = netdev_priv(dev); 350 if (test_bit(VIF_STATUS_CONNECTED, &vif->status)) 351 xenvif_up(vif); 352 netif_tx_start_all_queues(dev); 353 return 0; 354 } 355 356 static int xenvif_close(struct net_device *dev) 357 { 358 struct xenvif *vif = netdev_priv(dev); 359 if (test_bit(VIF_STATUS_CONNECTED, &vif->status)) 360 xenvif_down(vif); 361 netif_tx_stop_all_queues(dev); 362 return 0; 363 } 364 365 static int xenvif_change_mtu(struct net_device *dev, int mtu) 366 { 367 struct xenvif *vif = netdev_priv(dev); 368 int max = vif->can_sg ? ETH_MAX_MTU - VLAN_ETH_HLEN : ETH_DATA_LEN; 369 370 if (mtu > max) 371 return -EINVAL; 372 dev->mtu = mtu; 373 return 0; 374 } 375 376 static netdev_features_t xenvif_fix_features(struct net_device *dev, 377 netdev_features_t features) 378 { 379 struct xenvif *vif = netdev_priv(dev); 380 381 if (!vif->can_sg) 382 features &= ~NETIF_F_SG; 383 if (~(vif->gso_mask) & GSO_BIT(TCPV4)) 384 features &= ~NETIF_F_TSO; 385 if (~(vif->gso_mask) & GSO_BIT(TCPV6)) 386 features &= ~NETIF_F_TSO6; 387 if (!vif->ip_csum) 388 features &= ~NETIF_F_IP_CSUM; 389 if (!vif->ipv6_csum) 390 features &= ~NETIF_F_IPV6_CSUM; 391 392 return features; 393 } 394 395 static const struct xenvif_stat { 396 char name[ETH_GSTRING_LEN]; 397 u16 offset; 398 } xenvif_stats[] = { 399 { 400 "rx_gso_checksum_fixup", 401 offsetof(struct xenvif_stats, rx_gso_checksum_fixup) 402 }, 403 /* If (sent != success + fail), there are probably packets never 404 * freed up properly! 405 */ 406 { 407 "tx_zerocopy_sent", 408 offsetof(struct xenvif_stats, tx_zerocopy_sent), 409 }, 410 { 411 "tx_zerocopy_success", 412 offsetof(struct xenvif_stats, tx_zerocopy_success), 413 }, 414 { 415 "tx_zerocopy_fail", 416 offsetof(struct xenvif_stats, tx_zerocopy_fail) 417 }, 418 /* Number of packets exceeding MAX_SKB_FRAG slots. You should use 419 * a guest with the same MAX_SKB_FRAG 420 */ 421 { 422 "tx_frag_overflow", 423 offsetof(struct xenvif_stats, tx_frag_overflow) 424 }, 425 }; 426 427 static int xenvif_get_sset_count(struct net_device *dev, int string_set) 428 { 429 switch (string_set) { 430 case ETH_SS_STATS: 431 return ARRAY_SIZE(xenvif_stats); 432 default: 433 return -EINVAL; 434 } 435 } 436 437 static void xenvif_get_ethtool_stats(struct net_device *dev, 438 struct ethtool_stats *stats, u64 * data) 439 { 440 struct xenvif *vif = netdev_priv(dev); 441 unsigned int num_queues; 442 int i; 443 unsigned int queue_index; 444 445 rcu_read_lock(); 446 num_queues = READ_ONCE(vif->num_queues); 447 448 for (i = 0; i < ARRAY_SIZE(xenvif_stats); i++) { 449 unsigned long accum = 0; 450 for (queue_index = 0; queue_index < num_queues; ++queue_index) { 451 void *vif_stats = &vif->queues[queue_index].stats; 452 accum += *(unsigned long *)(vif_stats + xenvif_stats[i].offset); 453 } 454 data[i] = accum; 455 } 456 457 rcu_read_unlock(); 458 } 459 460 static void xenvif_get_strings(struct net_device *dev, u32 stringset, u8 * data) 461 { 462 int i; 463 464 switch (stringset) { 465 case ETH_SS_STATS: 466 for (i = 0; i < ARRAY_SIZE(xenvif_stats); i++) 467 memcpy(data + i * ETH_GSTRING_LEN, 468 xenvif_stats[i].name, ETH_GSTRING_LEN); 469 break; 470 } 471 } 472 473 static const struct ethtool_ops xenvif_ethtool_ops = { 474 .get_link = ethtool_op_get_link, 475 476 .get_sset_count = xenvif_get_sset_count, 477 .get_ethtool_stats = xenvif_get_ethtool_stats, 478 .get_strings = xenvif_get_strings, 479 }; 480 481 static const struct net_device_ops xenvif_netdev_ops = { 482 .ndo_select_queue = xenvif_select_queue, 483 .ndo_start_xmit = xenvif_start_xmit, 484 .ndo_get_stats = xenvif_get_stats, 485 .ndo_open = xenvif_open, 486 .ndo_stop = xenvif_close, 487 .ndo_change_mtu = xenvif_change_mtu, 488 .ndo_fix_features = xenvif_fix_features, 489 .ndo_set_mac_address = eth_mac_addr, 490 .ndo_validate_addr = eth_validate_addr, 491 }; 492 493 struct xenvif *xenvif_alloc(struct device *parent, domid_t domid, 494 unsigned int handle) 495 { 496 static const u8 dummy_addr[ETH_ALEN] = { 497 0xfe, 0xff, 0xff, 0xff, 0xff, 0xff, 498 }; 499 int err; 500 struct net_device *dev; 501 struct xenvif *vif; 502 char name[IFNAMSIZ] = {}; 503 504 snprintf(name, IFNAMSIZ - 1, "vif%u.%u", domid, handle); 505 /* Allocate a netdev with the max. supported number of queues. 506 * When the guest selects the desired number, it will be updated 507 * via netif_set_real_num_*_queues(). 508 */ 509 dev = alloc_netdev_mq(sizeof(struct xenvif), name, NET_NAME_UNKNOWN, 510 ether_setup, xenvif_max_queues); 511 if (dev == NULL) { 512 pr_warn("Could not allocate netdev for %s\n", name); 513 return ERR_PTR(-ENOMEM); 514 } 515 516 SET_NETDEV_DEV(dev, parent); 517 518 vif = netdev_priv(dev); 519 520 vif->domid = domid; 521 vif->handle = handle; 522 vif->can_sg = 1; 523 vif->ip_csum = 1; 524 vif->dev = dev; 525 vif->disabled = false; 526 vif->drain_timeout = msecs_to_jiffies(rx_drain_timeout_msecs); 527 vif->stall_timeout = msecs_to_jiffies(rx_stall_timeout_msecs); 528 529 /* Start out with no queues. */ 530 vif->queues = NULL; 531 vif->num_queues = 0; 532 533 vif->xdp_headroom = 0; 534 535 spin_lock_init(&vif->lock); 536 INIT_LIST_HEAD(&vif->fe_mcast_addr); 537 538 dev->netdev_ops = &xenvif_netdev_ops; 539 dev->hw_features = NETIF_F_SG | 540 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | 541 NETIF_F_TSO | NETIF_F_TSO6 | NETIF_F_FRAGLIST; 542 dev->features = dev->hw_features | NETIF_F_RXCSUM; 543 dev->ethtool_ops = &xenvif_ethtool_ops; 544 545 dev->tx_queue_len = XENVIF_QUEUE_LENGTH; 546 547 dev->min_mtu = ETH_MIN_MTU; 548 dev->max_mtu = ETH_MAX_MTU - VLAN_ETH_HLEN; 549 550 /* 551 * Initialise a dummy MAC address. We choose the numerically 552 * largest non-broadcast address to prevent the address getting 553 * stolen by an Ethernet bridge for STP purposes. 554 * (FE:FF:FF:FF:FF:FF) 555 */ 556 eth_hw_addr_set(dev, dummy_addr); 557 558 netif_carrier_off(dev); 559 560 err = register_netdev(dev); 561 if (err) { 562 netdev_warn(dev, "Could not register device: err=%d\n", err); 563 free_netdev(dev); 564 return ERR_PTR(err); 565 } 566 567 netdev_dbg(dev, "Successfully created xenvif\n"); 568 569 __module_get(THIS_MODULE); 570 571 return vif; 572 } 573 574 int xenvif_init_queue(struct xenvif_queue *queue) 575 { 576 int err, i; 577 578 queue->credit_bytes = queue->remaining_credit = ~0UL; 579 queue->credit_usec = 0UL; 580 timer_setup(&queue->credit_timeout, xenvif_tx_credit_callback, 0); 581 queue->credit_window_start = get_jiffies_64(); 582 583 queue->rx_queue_max = XENVIF_RX_QUEUE_BYTES; 584 585 skb_queue_head_init(&queue->rx_queue); 586 skb_queue_head_init(&queue->tx_queue); 587 588 queue->pending_cons = 0; 589 queue->pending_prod = MAX_PENDING_REQS; 590 for (i = 0; i < MAX_PENDING_REQS; ++i) 591 queue->pending_ring[i] = i; 592 593 spin_lock_init(&queue->callback_lock); 594 spin_lock_init(&queue->response_lock); 595 596 /* If ballooning is disabled, this will consume real memory, so you 597 * better enable it. The long term solution would be to use just a 598 * bunch of valid page descriptors, without dependency on ballooning 599 */ 600 err = gnttab_alloc_pages(MAX_PENDING_REQS, 601 queue->mmap_pages); 602 if (err) { 603 netdev_err(queue->vif->dev, "Could not reserve mmap_pages\n"); 604 return -ENOMEM; 605 } 606 607 for (i = 0; i < MAX_PENDING_REQS; i++) { 608 queue->pending_tx_info[i].callback_struct = (struct ubuf_info) 609 { .callback = xenvif_zerocopy_callback, 610 { { .ctx = NULL, 611 .desc = i } } }; 612 queue->grant_tx_handle[i] = NETBACK_INVALID_HANDLE; 613 } 614 615 return 0; 616 } 617 618 void xenvif_carrier_on(struct xenvif *vif) 619 { 620 rtnl_lock(); 621 if (!vif->can_sg && vif->dev->mtu > ETH_DATA_LEN) 622 dev_set_mtu(vif->dev, ETH_DATA_LEN); 623 netdev_update_features(vif->dev); 624 set_bit(VIF_STATUS_CONNECTED, &vif->status); 625 if (netif_running(vif->dev)) 626 xenvif_up(vif); 627 rtnl_unlock(); 628 } 629 630 int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref, 631 unsigned int evtchn) 632 { 633 struct net_device *dev = vif->dev; 634 struct xenbus_device *xendev = xenvif_to_xenbus_device(vif); 635 void *addr; 636 struct xen_netif_ctrl_sring *shared; 637 RING_IDX rsp_prod, req_prod; 638 int err; 639 640 err = xenbus_map_ring_valloc(xendev, &ring_ref, 1, &addr); 641 if (err) 642 goto err; 643 644 shared = (struct xen_netif_ctrl_sring *)addr; 645 rsp_prod = READ_ONCE(shared->rsp_prod); 646 req_prod = READ_ONCE(shared->req_prod); 647 648 BACK_RING_ATTACH(&vif->ctrl, shared, rsp_prod, XEN_PAGE_SIZE); 649 650 err = -EIO; 651 if (req_prod - rsp_prod > RING_SIZE(&vif->ctrl)) 652 goto err_unmap; 653 654 err = bind_interdomain_evtchn_to_irq_lateeoi(xendev, evtchn); 655 if (err < 0) 656 goto err_unmap; 657 658 vif->ctrl_irq = err; 659 660 xenvif_init_hash(vif); 661 662 err = request_threaded_irq(vif->ctrl_irq, NULL, xenvif_ctrl_irq_fn, 663 IRQF_ONESHOT, "xen-netback-ctrl", vif); 664 if (err) { 665 pr_warn("Could not setup irq handler for %s\n", dev->name); 666 goto err_deinit; 667 } 668 669 return 0; 670 671 err_deinit: 672 xenvif_deinit_hash(vif); 673 unbind_from_irqhandler(vif->ctrl_irq, vif); 674 vif->ctrl_irq = 0; 675 676 err_unmap: 677 xenbus_unmap_ring_vfree(xendev, vif->ctrl.sring); 678 vif->ctrl.sring = NULL; 679 680 err: 681 return err; 682 } 683 684 static void xenvif_disconnect_queue(struct xenvif_queue *queue) 685 { 686 if (queue->task) { 687 kthread_stop(queue->task); 688 put_task_struct(queue->task); 689 queue->task = NULL; 690 } 691 692 if (queue->dealloc_task) { 693 kthread_stop(queue->dealloc_task); 694 queue->dealloc_task = NULL; 695 } 696 697 if (queue->napi.poll) { 698 netif_napi_del(&queue->napi); 699 queue->napi.poll = NULL; 700 } 701 702 if (queue->tx_irq) { 703 unbind_from_irqhandler(queue->tx_irq, queue); 704 if (queue->tx_irq == queue->rx_irq) 705 queue->rx_irq = 0; 706 queue->tx_irq = 0; 707 } 708 709 if (queue->rx_irq) { 710 unbind_from_irqhandler(queue->rx_irq, queue); 711 queue->rx_irq = 0; 712 } 713 714 xenvif_unmap_frontend_data_rings(queue); 715 } 716 717 int xenvif_connect_data(struct xenvif_queue *queue, 718 unsigned long tx_ring_ref, 719 unsigned long rx_ring_ref, 720 unsigned int tx_evtchn, 721 unsigned int rx_evtchn) 722 { 723 struct xenbus_device *dev = xenvif_to_xenbus_device(queue->vif); 724 struct task_struct *task; 725 int err; 726 727 BUG_ON(queue->tx_irq); 728 BUG_ON(queue->task); 729 BUG_ON(queue->dealloc_task); 730 731 err = xenvif_map_frontend_data_rings(queue, tx_ring_ref, 732 rx_ring_ref); 733 if (err < 0) 734 goto err; 735 736 init_waitqueue_head(&queue->wq); 737 init_waitqueue_head(&queue->dealloc_wq); 738 atomic_set(&queue->inflight_packets, 0); 739 740 netif_napi_add(queue->vif->dev, &queue->napi, xenvif_poll, 741 NAPI_POLL_WEIGHT); 742 743 queue->stalled = true; 744 745 task = kthread_run(xenvif_kthread_guest_rx, queue, 746 "%s-guest-rx", queue->name); 747 if (IS_ERR(task)) 748 goto kthread_err; 749 queue->task = task; 750 /* 751 * Take a reference to the task in order to prevent it from being freed 752 * if the thread function returns before kthread_stop is called. 753 */ 754 get_task_struct(task); 755 756 task = kthread_run(xenvif_dealloc_kthread, queue, 757 "%s-dealloc", queue->name); 758 if (IS_ERR(task)) 759 goto kthread_err; 760 queue->dealloc_task = task; 761 762 if (tx_evtchn == rx_evtchn) { 763 /* feature-split-event-channels == 0 */ 764 err = bind_interdomain_evtchn_to_irqhandler_lateeoi( 765 dev, tx_evtchn, xenvif_interrupt, 0, 766 queue->name, queue); 767 if (err < 0) 768 goto err; 769 queue->tx_irq = queue->rx_irq = err; 770 disable_irq(queue->tx_irq); 771 } else { 772 /* feature-split-event-channels == 1 */ 773 snprintf(queue->tx_irq_name, sizeof(queue->tx_irq_name), 774 "%s-tx", queue->name); 775 err = bind_interdomain_evtchn_to_irqhandler_lateeoi( 776 dev, tx_evtchn, xenvif_tx_interrupt, 0, 777 queue->tx_irq_name, queue); 778 if (err < 0) 779 goto err; 780 queue->tx_irq = err; 781 disable_irq(queue->tx_irq); 782 783 snprintf(queue->rx_irq_name, sizeof(queue->rx_irq_name), 784 "%s-rx", queue->name); 785 err = bind_interdomain_evtchn_to_irqhandler_lateeoi( 786 dev, rx_evtchn, xenvif_rx_interrupt, 0, 787 queue->rx_irq_name, queue); 788 if (err < 0) 789 goto err; 790 queue->rx_irq = err; 791 disable_irq(queue->rx_irq); 792 } 793 794 return 0; 795 796 kthread_err: 797 pr_warn("Could not allocate kthread for %s\n", queue->name); 798 err = PTR_ERR(task); 799 err: 800 xenvif_disconnect_queue(queue); 801 return err; 802 } 803 804 void xenvif_carrier_off(struct xenvif *vif) 805 { 806 struct net_device *dev = vif->dev; 807 808 rtnl_lock(); 809 if (test_and_clear_bit(VIF_STATUS_CONNECTED, &vif->status)) { 810 netif_carrier_off(dev); /* discard queued packets */ 811 if (netif_running(dev)) 812 xenvif_down(vif); 813 } 814 rtnl_unlock(); 815 } 816 817 void xenvif_disconnect_data(struct xenvif *vif) 818 { 819 struct xenvif_queue *queue = NULL; 820 unsigned int num_queues = vif->num_queues; 821 unsigned int queue_index; 822 823 xenvif_carrier_off(vif); 824 825 for (queue_index = 0; queue_index < num_queues; ++queue_index) { 826 queue = &vif->queues[queue_index]; 827 828 xenvif_disconnect_queue(queue); 829 } 830 831 xenvif_mcast_addr_list_free(vif); 832 } 833 834 void xenvif_disconnect_ctrl(struct xenvif *vif) 835 { 836 if (vif->ctrl_irq) { 837 xenvif_deinit_hash(vif); 838 unbind_from_irqhandler(vif->ctrl_irq, vif); 839 vif->ctrl_irq = 0; 840 } 841 842 if (vif->ctrl.sring) { 843 xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif), 844 vif->ctrl.sring); 845 vif->ctrl.sring = NULL; 846 } 847 } 848 849 /* Reverse the relevant parts of xenvif_init_queue(). 850 * Used for queue teardown from xenvif_free(), and on the 851 * error handling paths in xenbus.c:connect(). 852 */ 853 void xenvif_deinit_queue(struct xenvif_queue *queue) 854 { 855 gnttab_free_pages(MAX_PENDING_REQS, queue->mmap_pages); 856 } 857 858 void xenvif_free(struct xenvif *vif) 859 { 860 struct xenvif_queue *queues = vif->queues; 861 unsigned int num_queues = vif->num_queues; 862 unsigned int queue_index; 863 864 unregister_netdev(vif->dev); 865 free_netdev(vif->dev); 866 867 for (queue_index = 0; queue_index < num_queues; ++queue_index) 868 xenvif_deinit_queue(&queues[queue_index]); 869 vfree(queues); 870 871 module_put(THIS_MODULE); 872 } 873